Not a week goes by without some model releases and this week ended with the long awaited Deepseek v4 in both Flash and Pro variants. I expect more and more inference providers to make this available soon as well as performance to increase as everybody tunes their stack. Other models of interest are GPT-5.5 and ChatGPT Images 2.0 from OpenAI which have received excellent reviews. Alibaba's dense smallish model Qwen 3.6-27B also surprised many with its excellent benchmarks
In other news, SpaceX's collaboration with Cursor with an option to buy them for $60B or pay them $10B for their work also generated buzz as well as this week there are a couple of news items related to legaltech that might be of interest to some readers
I also thought I'd share a LinkedIn post I made recently demonstrating an interesting way to watch Youtube via the Obsidian Web Clipper browser extension. I hope that readers find this useful
https://www.linkedin.com/feed/update/urn:li:activity:7453785426387107840/
Readers who think others in their family, friends and acquaintances who are curious in knowing more about rapidly evolving AI tools/services/use cases and would benefit from being subscribers of this weekly newsletter are encouraged to share this publication link to them and invite them to subscribe
The following messages were posted on the 'All Things AI' Telegram group from Sunday 19th Apr 2026 to Saturday 25th Apr 2026
Couple of xAI news including its Speech-to-Text API which I tested via some audio files from my conversation series and didn't get good results.
I've sent an email to their support to see if its something I did wrong or is the STT API more hype than real. Hopefully their devrels besides Elon will identify themselves on X
https://grok.com/release-notes/apr-17-2026
Instant, multi-speaker transcription across 25 languages - at the best price in the market.
xAI — Creators of Grok, the AI Chatbot
xAI builds Grok, an AI chatbot with voice chat, image and video generation, real-time search, and advanced reasoning. Try Grok at grok.com.
via Nils Phil
If you're in robotics, it's a really good idea to learn speak Chinese. I made this handy Claude tutor for you:
Simply set a Target and Aspiration, and ask Claude to reset your progress, then start playing the built in games.
Link:
GitHub - Broodsugar/mandarintutor: A set of markdown files to point Claude towards to get an...
A set of markdown files to point Claude towards to get an amazing mandarin tutor - Broodsugar/mandarintutor
via Vincent Alliaga
Vercel breached, citing Context.ai as the attack vector
Vercel April 2026 security incident | Vercel Knowledge Base
We've identified a security incident that involved unauthorized access to certain internal Vercel systems.
Joseph Cooper announced that he would personally DJing and vibe coding simultaneously on stage here in Hong Kong, April 30th, 2026
Nullshot Grounds HK 🇭🇰: Beats, Builders & AI with Ashford Benjamin · Luma
Welcome to the inaugural Nullshot Grounds. Hosted by Nullshot.ai in partnership with Ashford Benjamin, this is a brand new, quarterly networking series...
via Robby Yung
Moonshot AI Releases Kimi K2.6 with Top Open-Source Benchmarks
The Beijing-based company built K2.6 on a 1 trillion-parameter mixture-of-experts architecture, featuring a 256,000-token context window and native vision from a 400-million-parameter encoder. It scores highest among open-source models on tests like Humanity's Last Exam at 54.0, SWE-Bench Pro at 58.6, and others, even edging some closed models like GPT-5.4. Developers can grab the weights on Hugging Face under a commercial-friendly license, with real-world demos creating full React sites, 3D animations, and optimized inference engines from single prompts. Basecamp's David Heinemeier Hansson called the results 'incredible,' highlighting China's open-source push and big cost savings potential.
Open-source SOTA on HLE w/ tools (54.0), SWE-Bench Pro (58.6), SWE-bench Multilingual (76.7), BrowseComp (83.2), Toolathlon (50.0), Charxiv w/ python(86.7), Math Vision w/ python (93.2)What's new:
Long-horizon coding - 4,000+
Video hero sections - cinematic aesthetic, auto-composited WebGL shader animations - native GLSL / WGSL, liquid metal, caustics, raymarching Motion design - GSAP + Framer Motion
Microsoft is tightening rate limits on GitHub Copilot, removing Opus from $10-a-month subscriptions, and plans to move users to token/API-based billing later in 2026 in a sign that it's looking for way to cut costs for its AI services.
• Usage limits tightened; Pro+ offers 5X higher limits than Pro
Changes to GitHub Copilot plans for individuals - GitHub Changelog
As shared in our recent blog post, we’re making the following changes to Copilot plans for individuals as part of our ongoing efforts to ensure service reliability and a sustainable…
Exclusive: Microsoft To Shift GitHub Copilot Users To Token-Based Billing, Tighten Rate Limits
Executive Summary: * Internal documents reveal that Microsoft plans to temporarily suspend individual account signups to its GitHub Copilot coding product, as it transitions from requests (single interactions with Copilot) towards token-based billing.
OpenAI Launches Chronicle for Smarter Screen-Aware Coding in Codex
Chronicle, an experimental addition to the Codex macOS app, captures recent screen activity—like open docs, error messages, and projects—to build context-aware memories automatically. vailable to ChatGPT Plus and Pro users, it lets AI agents handle tasks such as debugging GitHub issues or drafting docs by scanning files, all without manual recaps. OpenAI leaders like Sam Altman and Greg Brockman call it 'surprisingly magical,' though it requires screen permissions, processes images on servers, and carries privacy risks like capturing sensitive info.
Today, we’re expanding the experiment with Chronicle, which improves memories using recent screen context.
Now, Codex can help with what you’ve been working on without you restating context.
Available for PRO subscriptions and on Mac to start. This is early and
Today, we’re expanding the experiment with Chronicle, which improves memories using recent screen context.
Now, Codex can help with what you’ve been working on without you restating context.
via Caterina Zhang
Hi everyone, I'm an independent musician, I do song writing and vocal, and use various Ai tools for final production. If anyone's interested in collaboration please let me know.
Here's the link of one piece of my work.
Recently came across Vibecon which is a virtual event with options for free and paid participation . Speakers from Deep mind, Microsoft, Kilocode, Monday.com , Cloudflare, together.ai
VIBECON - AI & LLM Conference 2026
Apr 29-30, Online. For those who build with AI. Speakers from Anthropic, GitHub, Replit, Cognition, Microsoft & more.
via Yat Siu
Interesting perspective here too if true haven't seen it for myself
From the SillyTavernAI community on Reddit
Explore this post and more from the SillyTavernAI community
New policy from Atlassian
Unless you opt out by August 17th 2026, data from Jira and Confluence will automatically be used for AI training. Some data cannot be opted out at all on some plans.
Unless you opt out by August 17th 2026, data from Jira and Confluence will automatically be used for AI training. Some data cannot be opted out at all on some plans.
x.com/kepano/status/…
via Valerii Anufriev
open-source reconstruction of Anthropic's Mythos based on publicly avaiiable data
GitHub - kyegomez/OpenMythos: A theoretical reconstruction of the Claude Mythos architecture, built from first principles using the available research literature.
A theoretical reconstruction of the Claude Mythos architecture, built from first principles using the available research literature. - kyegomez/OpenMythos
OpenAI Launches ChatGPT Images 2.0 with Smarter Visual Tools
OpenAI released ChatGPT Images 2.0 powered by the gpt-image-2 model that handles sharper edits, complex layouts, and real-time data like San Francisco's weather during a live demo. All users get the standard version, while paid plans unlock 'Thinking' mode for web searches, multiple variants, and consistent outputs up to 2K resolution. Benchmarks show it topping charts, with early testers creating intricate grids and realistic newspaper fronts that older tools couldn't match, opening doors for creators and developers alike.
Researchers behind the model explain. A thread:
Thinking & Intelligence in ChatGPT Images 2.0, demonstrated by @ayaanzhaque
The most capable image generation model yet, built for production-grade workflows with stronger text rendering, layout, editing, resolution, and multilingual rendering.
SpaceX announces partnership with Cursor AI including right to acquire for $60 billion or pay $10 billion
SpaceX revealed it is collaborating closely with Cursor AI to develop the world's best coding and knowledge work AI, leveraging Cursor's product for expert software engineers and SpaceX's Colossus supercomputer with a million H100 equivalents. Cursor has granted SpaceX the option to acquire the company later this year for $60 billion or pay $10 billion for their joint work.
The combination of Cursor’s leading product and distribution to expert software engineers with SpaceX’s million H100 equivalent Colossus training supercomputer will
Various links related to observations by many people of Anthropic removing Claude Code from its Pro subscription, subsequent clarification from an Anthropic staffer and then OpenAI responding to Anthropic
https://www.reddit.com/r/ClaudeAI/comments/1srzhd7/psa_claude_pro_no_longer_lists_claude_code_as_an/
https://www.reddit.com/r/Anthropic/comments/1rkijk6/anthropic_reduces_the_value_of_max_plans_by/
Pro users wanting it need Max now.
$100/month minimum. 5x jump.
I'm on Max 20x so I'm fine.
Flagging for anyone on Pro who's about to find out.
No announcement. Just a pricing page edit.
Transparency
Sullivan & Cromwell Files Emergency ‘Please Don’t Sanction Us For All These AI Hallucinations’ Letter
Spectrum: An open source framework that brings AI agents to iMessage, WhatsApp, Telegram, Slack, and SMS/RCS, through one unified API.
Introducing Spectrum: Agents for the rest of us - Photon Blog
Photon powers conversational AI agents on iMessage, WhatsApp, Telegram, and more. Open-source Spectrum framework + SDK for seamless agent interactions.
Couple of posts from Robby Yung
This is pretty insane:
→ Group in a private Discord guessed the endpoint from Anthropic's naming conventions
→ They figured out the
DESIGN.md lets you easily export and import your design rules from project to project. Instead of guessing intent, agents know
Qwen3.6-27B Alibaba Qwen team latest dense, open-source model, packing flagship-level coding power
🧠 Outstanding agentic coding — surpasses Qwen3.5-397B-A17B across all major coding benchmarks
💡 Strong reasoning across text & multimodal tasks
🔄 Supports thinking & non-thinking modes
Apache 2.0 — fully open, fully yours
Meet Qwen3.6-27B, our latest dense, open-source model, packing flagship-level coding power!Yes, 27B, and Qwen3.6-27B punches way above its weight.
What's new:
Outstanding agentic coding — surpasses Qwen3.5-397B-A17B across all major coding benchmarks
Strong
via OpenAI
Today we’re releasing OpenAI Privacy Filter, an open-weight model for detecting and redacting personally identifiable information (PII) in text.
Privacy Filter is a small model with frontier personal data detection capability. It is designed for high-throughput privacy workflows, and is able to perform context-aware detection of PII in unstructured text. It can run locally, which means that PII can be masked or redacted without leaving your machine. It processes long inputs efficiently, making redaction decisions in a quick, single pass.
The model is available today under the Apache 2.0 license on Hugging Face and Github. It is intended for experimentation, customization, and commercial deployment, and it can be fine-tuned for different data distributions and privacy policies.
via Anthony Estebe
Google’s infra announcements at Cloud Next'26 were substantial, not cosmetic:
They introduced 8th-gen TPUs with a split design: TPU 8t for training and TPU 8i for inference. Google says 8t delivers nearly 3x compute per pod vs Ironwood, while 8i connects 1,152 TPUs per pod for low-latency inference and high-throughput multi-agent workloads
TPU 8t achieves nearly three times the compute performance per pod over our previous generation, Ironwood.
TPU
OpenAI released GPT-5.5 to paid ChatGPT users on Plus, Pro, Business, and Enterprise plans, plus Codex, with a stronger Pro variant for higher tiers. It excels in agentic coding, research, data analysis, and multi-tool workflows, topping benchmarks like Terminal-Bench 2.0 at 82.7% and OSWorld-Verified at 78.7%. Early users praise its smarts but note the higher API costs of $5 per million input tokens and $30 per million output.
We've expanded browser use so Codex can interact with web apps, and test flows, click through pages, capture screenshots, and iterate on what it sees until it completes the
https://openai.com/index/introducing-gpt-5-5/
Based on Kevin Dent's question on how many in the group use ChatGPT vs Claude? Vano responded with
I use both actively. Ending up spending most Claude usage typically on Cowork. For coding, always using both and make them crosscheck each other's work.
For multi-tool agentic tasks such as monitoring Slack, updating tasks and statuses in project management tool, etc. ChatGPT is more efficient as quality is very much the same and much less daily/weekly usage allocation is spent compared to Claude.
And obviously some features are unique to each product. + using Gemini as part of Google Workspace
Came across OpenWork which bills itself as an open source, enterprise-ready alternative to Claude Cowork that companies can deploy internally — with their own LLMs, their own infrastructure, and their own compliance posture.
We launched 3 months ago, and already have 14k GitHub stars and 150k+ downloads.
OpenWork: The open source alternative to Claude Cowork | Y Combinator
The open source alternative to Claude Cowork. Founded in 2026 by Ben Shafii, OpenWork has 4 employees based in . OpenWork is hiring for 2 roles in engineering and design.
Roadmap - OpenWork Docs
What OpenWork supports today and what is coming next.
ComfyUI Raises $30 Million at $500 Million Valuation
The open-source tool, known for its node-based interface that gives pros precise control over generative AI workflows, secured $30 million led by Craft Ventures, pushing its total funding to $47 million. With over 4 million users, 60,000 community-built nodes, and 150,000 daily downloads—all without marketing—the funding will fuel Comfy Cloud for teams, better local installs, and support for new models while keeping the core open source. Founders like Yoland Yan and Robin Huang credit the community for the success, with studios already hiring 'ComfyUI artists' and investors praising its product-market fit.
ComfyUI raises $30M to scale open-source AI for creative production
Building the AI creation engine for visual professionals
Led by @craft_ventures , with @PaceCap , @chemistry , TruArrow, and others.
But before anything else: this belongs to the community.
ComfyUI started as one developer and one open-source repo. No
Below is my personal website which aggregates links to many of my socials as well as the various content and community that I curate. Feel free to share this link to others who you think may find this content/community useful to them
The cover image of this newsletter via generated via the OpenAI's ChatGPT Images 2.0 model within the Freepik tool via the following prompt
A breathtaking, sweeping desert landscape, suddenly transformed by a rare and miraculous rainfall, as millions of wildflowers of every shape, size, and color – velvety reds, sunshine yellows, delicate pinks, and soft lavenders – burst forth from the cracked, sandy earth, creating a striking visual contrast between the arid, weathered dunes and the vibrant, lush blossoms that blanket the terrain, with some flowers swaying gently in the breeze, while others stand tall and proud, as if defying the harsh, desert environment, under a bright blue sky with only a few wispy, white clouds.